One minor nit to pick. Audio in general tends to not be very small, but when all you care about is voice data, it can be compressed really really well: https://en.wikipedia.org/wiki/Speech_coding.
Still not âthatâ small. The lowest-bitrate codecs throw out a lot of information that is of utmost importance for fishing out commands from background noise. Then thereâs the issue of timing; speech recog is not as sensitive to latency as speech comm but it still has to be done in around a second or less to not be perceived as too laggy. All together, pretty easy to detect.
I guess Google is more concerned about competition from Apple, Microsoft and other tech companies rather than people getting paranoid over the use of this feature. It was something that Google had announced as a feature many months ago, when it was added to Chrome and few got upset about it, especially since it a feature that you had to opt-in to.
Once again, the code activating and deactivating the module is open source. So developers can clearly see when that module is being used by Chrome.
Slashdot was smart enough to see this as the garbage reporting that it is, and ignored it. Too bad Boingboing is a bit more gullible.
This⌠honestly the article asks for a hardware switch to disable camera/microphone. Iâd say youâre better off physically disabling them yourself (physically block the camera or disconnect the microphone).
I guess it depends what youâre comparing to, when talking about it being ânewâ. In terms of Chrome features, January 2014 is pretty ancient.
Unless someone finds a security bug that allows web content to flip the setting.
That never happens!
By âdisconnect,â you mean âcut traces on PCB with Xacto knife?â
Pretty much yes. Or desolder a part. Or pull a connector, if the microphone is attached by one. Alternatively, add a manual mechanical enable/disable switch. Same switch can be used for the power line of the camera module; this one is usually a USB camera, with an extra small/flat form factor to fit a laptop lid (and can be usually salvaged from dead laptops).
Hyperbole? Or honeypot? Look at how quickly the correct answer was posted.
Remember, kids. The government will try to argue with a straight face that they arenât violating your privacy rights when they subpoena, hack, or otherwise obtain data from a US corporation, regardless of how that data wound up in said corporationâs hands. As far as I can see, itâs all about plausible legal deniability, and google (nest, chrome), apple (siri etc), etc seem to be totally in on it; theyâre not just complicit, theyâre bending over backwards to achieve this.
Ok, Google, disable the OK Google feature.
This did prompt me to go find the âturn off the mikeâ function key combination on my laptop. Yes, itâs almost certainly implemented in software, but at least itâs a start.
I wish there was a way to flag the original post as âneeds an update, completely factually incorrectâ, since @doctorow doesnât read the comments on the BBS
Thatâs not exactly anyway an open API and Chrome tends to be one of the most secure out of all the browser doing a good job in sandboxing.
A more likely security breach if you go down that road is websites getting access to your microphone, as that is an actually open API for web audio. A site is supposed to get permission from the user, but if something were to break in security, a site would be more interested in that rather than sending data to Google for a voice search. Unless any other browsers have recently added the functionality, Chrome is the only browser that Iâm aware of which will mark a tab with a red record icon when a website is accessing your microphone or webcam.
This is my day job space (though for Firefox). I expect that they do less of a good job than the popular perception.
Iâd certainly wouldnât expect sandboxing to save folks here.
The point about an open API is well made though.
How can it listen for âOK Googleâ if itâs not always listening? Thatâs always been my issue with Google devices since the feature first rolled out. It doesnât matter if I have it on my phone, itâs impossible for me to tell whether or not anyone near me has it on theirs.
Any voice-controlled device has to keep listening.
If all the processing happens locally, no problem.
If it is a cloud monstrosity, to save the bandwidth and the associated costs (e.g. device power) the activation keyword is recognized locally (the single-word recognition is easier than full-scale recog) and only the followup is sent across the Net.
It can be compromised. But you can turn mostly anything with a microphone and a network connection into a listening device via a software update anyway.
This topic was automatically closed after 5 days. New replies are no longer allowed.